Are Hourly PM2.5 Forecasts Sufficiently Accurate to Plan Your Day? Individual Decision-Making in the Face of Increasing Wildfire Smoke

Abstract

Wildfire frequency is increasing as the climate changes, and the resulting air pollution poses health risks. Just as people routinely use hourly weather forecasts to plan their day’s activities around precipitation, reliable hourly air quality forecasts could help individuals reduce their exposure to air pollution. 

We evaluate six existing forecasts of ground-level fine particulate matter (PM_2.5) within the contiguous United States during the 2023 fire season from the perspective of individual decision-making. Forecasts use physical simulations as well as ensembling and artificial intelligence approaches. We ask whether these forecasts, potentially produced for other applications, might also reliably predict 1) whether to go outside on a day with potentially high PM_2.5 or 2) when to go outside for the lowest PM_2.5 exposure. We present multiple forecast skill metrics for the two tasks above and introduce a new evaluation metric for the task of deciding when to go outside. Time series plots are particularly effective in comparing forecasts during specific events. 

We find meaningful room for improvement in PM_2.5 forecasting skill, which might be realized by improving physical models, incorporating more data sources, and advancing artificial intelligence tools.

Significance Statement

Air pollution due to wildfires poses a large and increasing health challenge for individuals. Individuals, including those with preexisting respiratory conditions and outdoor recreationists, are increasingly interested in navigating personal decisions using air quality forecasts; these decisions include whether and when to go outside during a given day. Here, we outline a framework to evaluate how much individuals can trust existing forecasts for these personal decisions. Applying our framework, we find substantial room for improvement in existing forecasts. We outline development paths to better forecasts and suggest using our framework to benchmark future improvements in PM_2.5 forecasts.